Skip to content

Releases: giannimassi/wits

v0.1.0 — Initial release

21 Apr 17:08
@giannimassi giannimassi
v0.1.0
a50478f

Choose a tag to compare

View all tags
v0.1.0 — Initial release Latest
Latest

First tagged release of wits — composable reasoning skills for AI agents.

What's in v0.1.0

Three skills, installable independently or together:

think — 12 reasoning primitives

Structured thinking tools drawn from intelligence analysis (CIA SATs), cognitive science, philosophy, and decision theory. Each primitive is a self-contained protocol. Works on Claude Code, Codex CLI, and Gemini CLI.

Primitives: assumption-audit, disconfirmation, perspective-rotation, pre-mortem, hypothesis-generation, steelman, decomposition, confidence-calibration, causal-mapping, matrix-evaluation, base-rate-anchoring, inversion.

discuss — facilitated multi-agent discussion

Spawns a team of AI agents (facilitator, reasoning cartographer, critical lens, domain experts) to have a structured, time-limited discussion. Includes seed+react opening, stance-diversity enforcement, mid-discussion panel expansion, and NGT voting for convergence. Requires Claude Code (uses the Task tool).

recruit — expert persona registry

Create, cache, and reuse domain expert personas across sessions and skills. Silent caching — a failed or unavailable cache never blocks or prompts the user. Optional: set WITS_CACHE=off for stateless runs.

Install

npx skills add giannimassi/wits@skills/think
npx skills add giannimassi/wits@skills/discuss
npx skills add giannimassi/wits@skills/recruit

Notes

  • Follows the Agent Skills open standard
  • Persistent data goes to ~/.local/share/wits/ by default (override with WITS_DATA_DIR, disable with WITS_CACHE=off)
  • discuss + recruit work together: discuss assembles panels silently via recruit, then presents the roster once for user approval — no per-expert prompts on fresh installs
Assets 2
Loading